Hi,
The kmem extension of the memory cgroup is almost usable now. There is,
in fact, the only serious issue left: per memcg kmem caches may pin the
owner cgroup for indefinitely long. This is, because a slab cache may
keep empty slab pages in its private structures to optimize performance,
while we take a css reference per each charged kmem page.
The issue is only relevant to SLUB, because SLAB periodically reaps
empty slabs. This patch set fixes this issue for SLUB. For details,
please see patch 3.
Thanks,
Vladimir Davydov (3):
slub: don't fail kmem_cache_shrink if slab placement optimization
fails
slab: zap kmem_cache_shrink return value
slub: make dead caches discard free slabs immediately
include/linux/slab.h | 2 +-
mm/slab.c | 9 +++++++--
mm/slab.h | 2 +-
mm/slab_common.c | 21 +++++++++++++-------
mm/slob.c | 3 +--
mm/slub.c | 53 +++++++++++++++++++++++++++++++++++---------------
6 files changed, 61 insertions(+), 29 deletions(-)
--
1.7.10.4
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>